So, does R420 feature brilinear?
- Thread starter vb
- Start date
My understanding was that "Brilinear" was solely a nVidia card trait that used trilinear and bilinear mipmap transitions on the same texture stage (correct me if I'm wrong here). ATi does full trilinear on the first texture stage, but bilinear on the rest if you select Quality from the D3D Control Panel. However, this is not the definition of brilinear IIRC. ATi also allows full trilinear if you use ingame trilinear and AF, use third party tools like Radlinker and rTool, or simply modify the registry.
As far as I can tell, ATi has continued with their Quality control panel option in r420 but allow the user to set full trilinear through the methods above. On the flip side, nV doesn't offer said option save the 60.72 for nv40 and uses brilinear in 61.11 from recent articles.
[EDIT]
I misread the original question and thought it said brilinear in the r350 (somehow ignored the V like an idiot - small letter, BIG difference
). I then reread the linked thread and saw DaveB's findings. Thanks for the clarification and eradication of ignorance, DaveB.
As far as I can tell, ATi has continued with their Quality control panel option in r420 but allow the user to set full trilinear through the methods above. On the flip side, nV doesn't offer said option save the 60.72 for nv40 and uses brilinear in 61.11 from recent articles.
[EDIT]
I misread the original question and thought it said brilinear in the r350 (somehow ignored the V like an idiot - small letter, BIG difference
). I then reread the linked thread and saw DaveB's findings. Thanks for the clarification and eradication of ignorance, DaveB. 
It's not easy to tell, but I would say yes.
I made a post a while back in the Graphics Boards forum suggesting that 9600 owner should try mucking around with the texture slider to see the results. The reason for this is that ATI have put an adaptive texturing system in there - if you have the slider to the maximum position then the max quality from the application / control panel will be used, however if you move the texture quality slider down the texture properties will analysed and the best filtering (filtering type / brilinear mix / level of Ansio) for that texture will be utilised while maintaining as much performance as possible; the further down the slider you go the more aggressive the selection is towards performance over IQ.
Now, if you use the texture slider on the texture filtering test then it just shows maximum filtering, probably due to this algorithm. I believe that R420 will utilise this to the maximum, which also means the bilinear / trilinear mix would be there as well.
I made a post a while back in the Graphics Boards forum suggesting that 9600 owner should try mucking around with the texture slider to see the results. The reason for this is that ATI have put an adaptive texturing system in there - if you have the slider to the maximum position then the max quality from the application / control panel will be used, however if you move the texture quality slider down the texture properties will analysed and the best filtering (filtering type / brilinear mix / level of Ansio) for that texture will be utilised while maintaining as much performance as possible; the further down the slider you go the more aggressive the selection is towards performance over IQ.
Now, if you use the texture slider on the texture filtering test then it just shows maximum filtering, probably due to this algorithm. I believe that R420 will utilise this to the maximum, which also means the bilinear / trilinear mix would be there as well.
So using FilterTest (or similar) and looking at texture stages is no longer a decent way to get a quick overview of overall image quality on any given card/driver combo, since the driver/hardware may be 'optimising' for FilterTest's simple texture usage?
Rys
Rys
Mintmaster
Veteran
Just noticed something in Tom's review:
Looks like brilinear is there, should ATI decide to use it. Anyone think ATI is currently using it in some circumstances? Their anisotropic filtering performance is unbelievable. Too bad they didn't fix that angular dependence, because I doubt it would have cost much performance.Tom's Hardware said:
AlphaWolf said:
The point is you get no control over filtering quality. Force CP AF, get bilinear on all texture stages bar the first, regardless of whether the app asked for trilinear or not. He's just stating a perfectly valid fact.
That applications, games or otherwise, don't provide the facility to change AF is also a problem and also a fact. Don't get aniso control in your game? You'll force it in the CP. And what do you end up with? Worse quality filtering than ATI could offer via a simple CP toggle button.
All Chalnoth is doing is stating facts, facts that need considering.
Rys
DaveBaumann said:
how long has the adaptive filtering been in?
I posted quite a while ago about a bilinear/trilinear mix of filtering that the 8500 was using and that in opengl (tex pref at anything but high quality)as soon as coloured mipmaps were turned on it would switch up to full trilinear but when they were off it would run in it's crude version of brilinear.
A few driver releases. I don't think it would be enabled for < R300 products, or it may even be RV3x0 only.
Mintmaster said:
I'd wager the rather large texture fill-rate (in relation to bandwidth per pixel) is likely to have something to do with this - we are talking over 8G bilinear samples per second.
Similar threads
